Stockfish testing with multi-threaded WebAssembly

I actually have an analysis tool that can store evals for any position irrespective of depth. Also multipv is not limited, so you can store the evaluation of all moves for a position. The engine runs in the cloud. Only problem is of course that there is only one engine instance, so only one user can analyze at a time. The evaluations are stored globally, not per user.

Visit:

fbserv.herokuapp.com/

Go to the "Board" tab and press "Analyze". To store the analysis press "Store Stop".

In short: The limitations are intended to balance database size and utility. Unbalanced positions are less likely to be reached on a regular basis. If there's none or only low depth cloud analysis, then the client will just redo it. It works quite well, but the exact cutoffs are essentially arbitrary.

My 2 cents on WASMX, for me here (chrome on OSX, same processors setting as PNACL) the speed is pretty much in the same order of magnitude, and in some moments it even reaches higher peaks. So for me i'ts a keeper.

@spaitz

Is there any way to make analysis board use other than WASMX?

I have memories of how PNACL performed, but cannot make a direct comparison because I cannot set it back to PNACL. If my memories are correct WASMX is still half as performant in kNodes/sec.

@lishadowapps No idea.

I know that when WASM (no x) was tested, for me it was like 5 times slower, it popped up often and it meant local analysis was unusable. That's when I started noticing knodes per second and that's why I know WASMX is up there if not faster.
I could get rid of WASM by cleaning cookies.